JP5166909B2 - Inspection device and inspection method - Google Patents

Description

  The present invention relates to an inspection apparatus and an inspection method, and more particularly to an inspection apparatus, an inspection method, and a program using an X-ray image.

When a plurality of tubes are arranged in a skirt shape and brazed between them, it is necessary to inspect whether or not there are defects (hereinafter referred to as “voids”) inside the brazed portion. Conventionally, various methods have been applied to inspect the brazed portion.
Japanese Patent No. 3530478 (Patent Document 1) shows a technique using X-rays for inspection of a brazing portion. According to this method, a predetermined part is imaged a plurality of times while gradually changing the X-ray intensity, and the density of the obtained X-ray transmission image is superimposed. The density of the superimposed X-ray transmission image is different between pixels, and the difference is further calculated from the obtained difference density to obtain an inflection point. A closed region is created by connecting the inflection points, and a region in which no other closed region exists is extracted.

In relation to the above description, an X-ray fluoroscopic inspection apparatus is shown in Japanese Patent Laid-Open No. 9-196865 (Patent Document 2). In this conventional example, the object to be inspected is irradiated with X-rays, and contamination of foreign matters is examined.
Japanese Patent No. 3530478 JP-A-9-196865

An object of the present invention is to provide an inspection method capable of efficiently inspecting a subject by performing image processing based on the shape of a void, a program therefor, and an inspection apparatus therefor.
Another object of the present invention is to provide an inspection method capable of efficiently inspecting a subject by performing image processing based on the feature amount even if void images are connected, a program therefor, and a program therefor It is to provide an inspection apparatus.
Still another object of the present invention is to provide an inspection method capable of efficiently acquiring only an image of a void in an inspection target region, a program therefor, and an inspection apparatus therefor.

  The means for solving the problem will be described below using the numbers and symbols used in the [Embodiments of the Invention]. These numbers and symbols are added to clarify the correspondence between the description of [Claims] and the description of the embodiments of the invention, but are described in [Claims]. It should not be used to interpret the technical scope of the invention.

In one aspect of the present invention, an examination method includes (a) a step of capturing an X-ray image of a subject, for example, and (b) detection candidates separated from each other based on a first luminance from the X-ray image. Generating an image; (c) selecting a first void candidate image based on an aspect ratio of each of the detection candidate images; and (d) each of the remaining detection candidate images based on a second luminance. Generating a second candidate void image, and (e) synthesizing the first void candidate image and the second void candidate image to generate a void candidate image.
Thereby, a void can be detected efficiently.

The inspection method may further include a step of generating a void detection image indicating the detected void from the void candidate image based on the reference feature amount of the void.
The step (f) includes: (f1) a step of determining the image as a void detection image when the filling rate of the void candidate image is higher than a reference filling rate as the reference feature amount; and (f2) the void. Dividing the void candidate image into void candidate element images when the filling rate of the candidate images is equal to or less than the reference filling rate; (f3) calculating a shape feature amount of the void candidate element image; (f4) It is preferable that the method includes determining the void candidate element image as a void detection image when the shape feature amount satisfies a void criterion.
The inspection method may further include (g) a step of excluding those existing in a predetermined region corresponding to a part of the subject from the void candidate image.
The excluding step includes a step of detecting an edge of the component from an image obtained from the binarization processing of the X-ray image, a step of setting a straight line region from the detected edge, and a step based on the straight line region. And masking a part of the determined void detection image.
The step (a) includes a step of controlling the positions of the X-ray generator and the X-ray camera so that the specific part of the subject becomes the center of the X-ray image when imaging different parts of the subject. It is preferable to comprise.

  In another aspect of the present invention, a computer-readable program for realizing any of the above inspection methods is provided.

  In another aspect of the present invention, the inspection apparatus moves in response to the first movement instruction, generates an X-ray in response to the generation instruction, and responds to the subject movement instruction. An X-ray camera (8) for imaging the subject using an X-ray generated from the X-ray generator as a second movement instruction, and an X-ray camera (8) for moving the position and orientation of the subject. An image capture (10) for generating an X-ray image from an imaging result of a line camera, a display device (12), the generation instruction is output to the X-ray generator, the actuator is driven, and the image capture And a processing unit (2) for acquiring a detection target image from the X-ray image. The processing unit (2) issues the first and second movement instructions and the subject movement instruction, issues the generation instruction to the X-ray generator, and captures an X-ray image of the subject. , Generating detection candidate images separated from each other based on a first luminance from the X-ray image, selecting an image of a first void candidate based on an aspect ratio of each of the detection candidate images, and setting a second luminance Based on each of the remaining detection candidate images, a second void candidate image is generated, and the first void candidate image and the second void candidate image are synthesized to generate a void candidate image and output to the display device To do.

The processing unit (2) preferably generates a void detection image indicating the detected void from the void candidate image based on the reference feature amount of the void.
When the filling rate of the candidate void image is higher than a reference filling rate as the reference feature amount, the processing unit (2) determines the image as a void detection image, and the filling rate of the void candidate image is the reference value When the filling factor is less than or equal to the filling factor, the void candidate image is divided into void candidate element images, the shape feature amount of the void candidate element image is calculated, and the void candidate element image is voided when the shape feature amount satisfies a void criterion. It is preferable to determine the detected image.
The processing unit (2) preferably excludes those present in a predetermined region corresponding to a part of the subject from the void candidate image. At this time, the processing unit (2) detects an edge of the component from the image obtained from the binarization processing of the X-ray image, sets a straight line region from the detected edge, and sets the straight line region to the straight line region. Preferably, a part of the void detection image determined based on the mask is masked.
The processing unit (2) sets the position of the set of the X-ray generator and the X-ray camera so that the specific part of the subject becomes the center of the X-ray image when imaging different parts of the subject. It is preferable to control.

According to the present invention, utilizing the fact that the brazing unit is dark on the X-ray transmission image, the center of the brazing unit is automatically extracted by image processing, and the amount of deviation from the center of the brazing unit on the image And the imaging mechanism is positioned (adjusted).
Further, the X-ray transmission image is binarized with a low threshold value, and is divided into an area that is considered a void based on the shape characteristics (circularity and size) of the extracted area, and an area based on the structure of the brazing portion. Thereafter, the luminance of the latter area is examined in the area and binarized again with a higher threshold value. In this way, a void detection image can be obtained efficiently.
Furthermore, determination is made based on the extracted features of the void area (filling rate in the circumscribed area, etc.), and the image is separated. Thus, voids can be detected more accurately.
In addition, a tube-shaped region is obtained from the contour of the brazing portion (the contour of the upper portion) and is used to mask the void detection image. In this way, a more appropriate void detection image can be obtained.

Hereinafter, an inspection apparatus according to the present invention will be described in detail with reference to the accompanying drawings.
In this embodiment, the subject is a skirt portion of a rocket engine, for example, and a large number of tubes are arranged so as to form a side surface of the engine skirt, and the tubes are brazed. For this reason, as shown in FIG. 8A, the interval between the pipes is narrow at the upper part of the engine skirt, and the interval between the pipes is wide at the lower part. In the inspection apparatus according to the first embodiment of the present invention, the engine skirt is placed not in the vertical direction but in the horizontal direction, and the brazing portion is inspected.

FIG. 1 is a block diagram showing the configuration of the inspection apparatus according to the first embodiment of the present invention. Referring to FIG. 1, the inspection apparatus includes a processing unit 2, an X-ray generator 4, an actuator 6 that drives a subject 20, an X-ray camera 8, an image capture 10, and a display unit 12. The processing unit 2 has a storage unit (not shown).
The processing unit 2 outputs a movement instruction to the actuator 6. The actuator 6 moves the subject in response to the movement instruction and controls the position, orientation, etc. of the subject. Further, the processing unit 2 issues a movement instruction to the X-ray generator 4. In response to the movement instruction, the X-ray generator 4 moves to change the imaging position. The position of the X-ray camera 8 is also moved corresponding to the position of the X-ray generator 4. The processing unit 2 controls the image capture 10 to capture an X-ray transmission image of the subject imaged by the X-ray camera 8. The processing unit 2 performs image processing on the X-ray transmission image, extracts a defect of the subject, and outputs the result to the display unit 12.

Next, the operation of the inspection apparatus of the present invention will be described with reference to FIG.
The processing unit 2 issues a movement instruction to the X-ray generator 4 and the X-ray camera 6 to move the X-ray generator 4 and the X-ray camera 6 to the examination site. Next, the processing unit 2 issues a generation instruction to the X-ray generator 4 and an imaging instruction to the X-ray camera 6 to image an inspection target portion of the engine skirt as a subject. Under the processing of the processing unit 2, the imaging result is sent to the image capture 10. The image capture 10 generates an X-ray transmission image of the subject shown in FIG. 3A from the imaging result and transfers it to the processing unit 2. When receiving the X-ray transmission image, the processing unit 2 temporarily stores it in a storage unit (not shown). In the X-ray original image shown in FIG. 3A, the upper part and the lower part correspond to the tube, and the center part corresponds to the brazing part. This is because the image was taken with the engine skirt sideways. In the center portion, a horizontally long area and an elliptical area are shown in the lower right due to uneven thickness of the brazed portion, and an elliptical area is further shown in the horizontally long area. These elliptical areas are voids to be detected by the inspection apparatus of the present invention.

  The processing unit 2 reads out the X-ray transmission image and performs binarization processing to generate an inspection region extraction binary image. The inspection area generally corresponds to the row portion. The inspection region extraction binary image is an image in which a non-inspection region and an inspection region are divided by binary data. For example, the upper and lower portions of the X-ray transmission image as non-inspection regions correspond to “0”, and the central portion as the inspection region corresponds to “1”. The generated inspection region extraction binary image is stored in the storage unit.

In the processing of the X-ray transmission image, in step S2, the processing unit 2 performs a morphology process (pre-processing) for the purpose of noise removal on the X-ray original image.
Next, in step S4, a lower threshold is set from the input unit. This threshold value may be set by the user or automatically set by the processing unit 2. The processing unit 2 performs binarization processing of the X-ray original image based on the set threshold value. Thereby, the binarized image shown in FIG. 3B is obtained. If binarization using a lower threshold is adopted, void candidates can be extracted without exception.

  In step S6, the processing unit 2 reads the inspection area extraction binary image from the storage unit, and calculates a logical product (AND) in units of pixels from the image binarized in step S4. In this way, as shown in FIG. 3B, it is possible to delete an image of the inspection region corresponding to the brazing portion by deleting the region other than the inspection region from the X-ray original image. Subsequently, in step S8, the processing unit 2 gives a label to each of the images of the high luminance area existing in the inspection area.

  In step S10, the processing unit 2 determines whether or not the label image to which the label is attached is a horizontally long label image. In general, the void is assumed to be a substantially circular shape or a substantially elliptical shape with a small (horizontal / vertical) ratio or close to 1. An ellipse having a large (horizontal / vertical) ratio or a large (vertical / horizontal) ratio is not assumed to be a void. Further, even if the label image is a substantially circular or substantially oval image, it is considered that an image having a large radius or height and a large image area is not a void. Therefore, as shown in FIG. 3C, when the label image is a horizontally long label, that is, when the label image is an ellipse with a large (horizontal / vertical) ratio, step S12 is executed. As shown in FIG. 3G, when it is not a landscape label, step S30 is executed. In step S30, as shown in FIG. 3H, the label image is a void based on the shape characteristics of the label image, for example, the circularity of whether the (horizontal / vertical) ratio is close to 1, and the size of the label image. It is determined whether or not the image can be considered. When it is determined that the label image is a void image, step S34 is executed, and the label image shown in FIG. 3 (i) is stored in the storage unit as a void candidate image (1). Further, when the label image is not determined to be a large image or the like, step S32 is executed, and the processing unit 2 sets a higher value as the threshold value. Thereafter, step S4 is executed except for the threshold setting process. In this way, a binarized image from which noise having higher luminance than before is removed is obtained. Then, the process after step S6 is performed.

  The horizontally long label image is not a void image, but may be a structure-dependent image extracted by the thickness distribution of the brazing portion. Therefore, the processing unit 2 executes the processing from step S12 to step S18 for each of the horizontally long labels. In step S14, the processing unit 2 executes a binarization process based on a threshold set for each label. When the horizontally long label image is based on the thickness distribution of the brazed portion, the brightness of the horizontally long label image is generally low. On the other hand, since the brightness of the void image is relatively high, the void image can be separated by increasing the threshold value. In step S16, a new label is assigned to each of the newly obtained images. Thereafter, as shown in FIG. 3D, the label image is determined to be a void image based on the shape characteristics of the label image, for example, the circularity of whether the (horizontal / vertical) ratio is close to 1 or the size of the label image. It is determined whether it can be considered. When it is determined that the label image is a void image, the image shown in FIG. 3E is stored in the storage unit as a void candidate image (2) in step S20.

Next, in step S22, the void candidate image (1) obtained in step S34 and the void candidate image (2) obtained in step S20 are read from the storage unit, synthesized, and shown in FIG. 3 (f). As shown, a new void detection image showing the detected void is generated. Each of the void images in the void detection image is given a label in step S24.
In the conventional method, in the extraction of the void, a portion where the luminance value is greatly changed by the brightness processing of the brazing portion is extracted as an edge, and the edge is combined in the vicinity to determine the void portion. At that time, since the feature of the general shape (circle / ellipse) of the void is not used, for example, even in a dark part as a whole, if there is a luminance difference, it is extracted as an edge. For this reason, there was a case where it does not coincide with human inspection of finding a bright part locally. In addition, there is a method of simply extracting a bright portion above a certain level, but the entire linear bright portion due to the structure of the brazing portion is extracted, resulting in excessive extraction.
In the present embodiment, the image is first binarized with a low threshold, and the region (region 2) that is considered to be a void based on the extracted shape features (circularity and size) of the eyelids, and the structure of the brazing portion (region 1). ). The latter is further binarized at a higher threshold by examining the luminance distribution within that region. Thereby, a void detection image is detected based on a more appropriate luminance.

  Next, in step S40, a void is detected based on the feature of the void image (label image) to which the label is assigned. Thereafter, in step S50, a process is performed to remove those detected as voids at a position corresponding to the tube portion. Note that steps S24, S40, and S50 may be omitted.

  Next, the configuration of the inspection apparatus according to the second embodiment of the present invention is the same as that of the inspection apparatus according to the first embodiment. Therefore, the description is omitted. The operation of the inspection apparatus according to the second embodiment will be described with reference to FIG. This operation corresponds to step S40 in the first embodiment.

  Conventionally, as shown in FIG. 5 (a), when there is a connected image, as shown in FIG. 5 (b), T1 and T2 are the distance of the image from the tube (junction width), When T3 is a brazing part width, the non-joint part width is obtained by T3- (T1 + T2), and the joining rate is obtained by {T3- (T1 + T2)} / T3. Conventionally, as a result of the obtained joint ratio, the void image may be excessively rejected.

  In this embodiment, first, a label image is obtained by assigning a label to each of the elements of the image. This step corresponds to step S24 of the first embodiment. In step S41, as shown in FIG. 5C, the processing unit 2 calculates each label feature amount (filling rate) of the label image. As shown in FIG. 5D, the filling rate means {(extracted void area) / (area of circumscribed rectangle)}. In FIG. 5C, the ratio between the area of the dark part and the area of the rectangle adjacent to the dark part outside is calculated. In step S42, it is determined whether or not the filling rate is equal to or less than a reference value. If it is determined that the filling rate is not below the reference value, step S46 is executed.

  If it is determined in step S42 that the filling rate is equal to or less than the reference value, the image is determined to be an image in which a plurality of void elements are combined, and step S43 is executed. In step S43, contraction / expansion processing is performed on the target label image. In the contraction process, a certain width around the label image as the target image, for example, one pixel or two pixels is deleted. That is, the luminance is set to zero. When there is a request for further shrinkage, the image is deleted from the target image by a predetermined width. The brightness of pixels to be deleted varies, and they are set to 0 brightness. Thus, as shown in FIG. 5E, a contracted image having the separated images is obtained. Subsequently, an expansion process is performed in order to approximate the original image. In the expansion process, a certain width, for example, a width corresponding to one pixel is added around the target image. Further, when there is a demand for expansion, a predetermined width is added around the target image. The luminance of the added pixel is constant. In this way, an expanded image shown in FIG. 5F is obtained. In this way, by applying the contraction / expansion processing to the label image, one label image can be decomposed into three appropriate images.

In step S44, the processing unit 2 calculates a label feature amount for each of the images obtained in step S43 as shown in FIG. The feature amount is calculated as {T3− (T1 + T2 + T4)} / T3. The definition of the values T1 to T4 is shown in FIG. That is, T1 is the distance from the upper tube portion of the upper image, T2 is the distance from the lower tube portion of the lower image, T3 is the distance between the tube portions, and T4 is the separated image. Is the distance between.
In step S45, the processing unit 2 determines whether or not the calculated feature amount satisfies the void determination criterion. When the void determination criterion is not satisfied, step S47 is executed, and it is determined that the determination target image is a non-void image. On the other hand, when the void determination criterion is satisfied, step S46 is executed. In step S46, the target image is determined as a void image. Further, similarly to the determinations in steps S8 and S16 in the first embodiment, it is determined whether or not each target image is a horizontally long image, and further whether or not it is a void image is determined based on the circularity or the like. .

In the past, whatever the shape of the connected parts was recognized as a void. For this reason, when a plurality of voids are connected, it was mistakenly recognized as a large void as a whole. In the present embodiment, whether or not the image is a single void image is determined based on the extracted void area characteristics (filling ratio = extracted void area / circumscribed rectangle area), and is determined to be a connected void image. When it is separated, it is determined whether or not each separated image is a void based on the features (size, circularity, etc.) of the void shape. As a result, the void can be determined more precisely.
Note that the second embodiment may be executed after the first embodiment.

  Next, the configuration of the inspection apparatus according to the third embodiment of the present invention is the same as that of the inspection apparatus according to the first embodiment. Therefore, the description is omitted. The operation of the inspection apparatus according to the third embodiment will be described with reference to FIG. This operation corresponds to step S50 of the operation of the first embodiment.

  Conventionally, as shown in FIG. 7A, the portion where the wax protrudes above the tube portion is also recognized as a dark brazing portion, and has been an analysis target. For this reason, voids that do not require extraction have also been extracted, and the inspection accuracy has been reduced.

In the present embodiment, the brazed portion extraction binarized image described in the first embodiment is prepared. This binarized image is the same as the inspection target region image in the first embodiment.
Next, in step S52, the processing unit 2 extracts an edge from the row extraction binary image. Subsequently, in step S54, a straight line representing the pipe portion is extracted using the extracted edge. In step S56, the processing unit 2 uses the extracted straight line to generate a binarized image representing the tube region as shown in FIG. 7B. In this binarized tube portion image, the non-tube portion region and the tube portion region are images that are divided by binary data. For example, in the non-tube region, the pixel corresponds to a luminance of “1”, and the luminance of the tube region corresponds to “0”. In step S58, the logical product (AND) of the binarized tube portion image and the void detection image is calculated for each pixel. As a result, the portion where the brazing shown in FIG. 7C protrudes from the tube portion is excluded from the inspection symmetry, and the accuracy of the void inspection is improved.

  Conventionally, in determining a processing target area, a dark portion is recognized as a low portion, but a portion where the low portion protrudes is also an object of analysis, and unnecessary voids are extracted to reduce inspection accuracy. On the other hand, in the third embodiment of the present invention, fitting is performed from the low part outline (dark outline) to the tube shape, and the region is excluded as the tube portion.

Next, the configuration of the inspection apparatus according to the fourth embodiment of the present invention is the same as that of the inspection apparatus according to the first embodiment. Various objects are inspected in the present invention, and one of the inspected objects is a rocket engine skirt. In the skirt portion, a large number of tubes are arranged so as to form the side surface of the engine skirt. For this reason, as shown in FIG. 8A, the interval between the pipes is narrow at the upper part of the engine skirt, and the interval between the pipes is wide at the lower part. When the engine is placed sideways, it is narrow on one side and wide on the other side. Therefore, as shown in FIG. 8B, when inspecting various parts of the engine skirt, the X-ray generator 4 and the X-ray camera 8 are inspected immediately before by simply moving them in the horizontal direction. There is a case where the part corresponding to the broken part does not come to the center of the image. For this reason, in the fourth embodiment, the X-ray generator 4 and the X-ray camera 8 are moved while maintaining the continuity of the images so that the tube portion of the examination target portion is always at the center of the image.
As described above, in this embodiment, the brazing unit is dark so that the center of the brazing unit is automatically extracted by image processing, and the amount of deviation from the center of the image at the center of the brazing unit is obtained to obtain an X-ray. The positions of the generator 4 and the X-ray camera 8 are determined and controlled.

As described above, according to the present invention, by utilizing the fact that the brazing part is dark, the center of the brazing part is automatically extracted by image processing, and the amount of deviation from the center of the brazing part on the image is determined. And the imaging mechanism is positioned (adjusted).
Further, the X-ray transmission image is binarized with a low threshold value, and is divided into a region considered as a void based on the extracted shape characteristics (circular gradient, size), and a region due to the structure of the brazing portion. Thereafter, the luminance of the latter area is examined in the area and binarized again with a higher threshold value. In this way, a void detection image can be obtained efficiently.
Furthermore, determination is made based on the extracted features of the void area (filling rate in the circumscribed area, etc.), and the image is separated. Thus, voids can be detected more accurately.
In addition, a tube-shaped region is obtained from the contour of the brazing portion (the contour of the upper portion) and is used to mask the void detection image. In this way, a more appropriate void detection image can be obtained.

FIG. 1 is a diagram showing the configuration of the inspection apparatus according to the first embodiment of the present invention. FIG. 2 is a flowchart for explaining the operation of the inspection apparatus according to the first embodiment of the present invention. FIG. 3 is a diagram for explaining the operation of the inspection apparatus according to the first embodiment of the present invention. FIG. 4 is a flowchart for explaining the operation of the inspection apparatus according to the second embodiment of the present invention. FIG. 5 is a diagram for explaining the operation of the inspection apparatus according to the second embodiment of the present invention. FIG. 6 is a flowchart for explaining the operation of the inspection apparatus according to the third embodiment of the present invention. FIG. 7 is a diagram for explaining the operation of the inspection apparatus according to the third embodiment of the present invention. FIG. 8 is a diagram showing a state in which the subject is irradiated with X-rays in the present invention.

Explanation of symbols

2: Processing unit 4: X-ray generator 6: Actuator 8: X-ray camera 10: Image capture 12: Display device 20: Subject

Claims (9)

(A) capturing an X-ray image of the subject;
(B) generating detection candidate images separated from each other based on a first luminance from the X-ray image;
(C) selecting an image of a first void candidate based on an aspect ratio of each of the detection candidate images;
(D) generating a second void candidate image from each of the remaining detection candidate images based on the second luminance;
(E) synthesizing the first void candidate image and the second void candidate image to generate a void candidate image ;
(F) generating a void detection image indicating a detected void from the void candidate image based on a reference feature amount of the void ; and
The step (f)
(F1) When the filling rate of the void candidate image is higher than the reference filling rate as the reference feature quantity, the step of determining the image as a void detection image, where the filling rate is {(void candidate image Area) / (area of the circumscribed rectangle of the void candidate image)},
(F2) dividing the void candidate image into void candidate element images when the filling rate of the void candidate image is equal to or less than the reference filling rate;
(F3) calculating a shape feature amount of the void candidate element image;
(F4) determining that the void candidate element image is a void detection image when the shape feature value satisfies a void criterion;
An inspection method comprising : The inspection method according to claim 1,
(G) The inspection method further comprising the step of excluding from the void candidate images those existing in a predetermined region corresponding to some constituents of the subject. The inspection method according to claim 2 ,
The excluding step includes:
Detecting an edge of the component from an image obtained from the binarization processing of the X-ray image;
Setting a straight line region from the detected edges;
Masking a part of the void detection image determined based on the straight line region. In the inspection method according to any one of claims 1 to 3,
The step (a) includes:
An inspection method comprising a step of controlling the positions of an X-ray generator and an X-ray camera so that when a different part of the subject is imaged, the specific part of the subject becomes the center of the X-ray image. A computer-readable program for realizing the inspection method according to any one of claims 1 to 4 . An X-ray generator that moves in response to the first movement instruction and generates X-rays in response to the generation instruction;
An actuator that moves the position and posture of the subject in response to the subject movement instruction;
An X-ray camera that images the subject using X-rays generated from the X-ray generator as a second movement instruction;
Image capture for generating an X-ray image from the imaging result of the X-ray camera;
A display device;
A processing unit that outputs the generation instruction to the X-ray generator, drives the actuator, and acquires a detection target image from the X-ray image from the image capture;
The processing unit issues the first and second movement instructions and the subject movement instruction, issues the generation instruction to the X-ray generator to capture an X-ray image of the subject, and From the line image, detection candidate images separated from each other based on the first luminance are generated, and the first void candidate image is selected based on the aspect ratio of each of the detection candidate images, and the remaining based on the second luminance is selected. Generating a second void candidate image from each of the detected candidate images, generating the void candidate image by synthesizing the first void candidate image and the second void candidate image, and outputting to the display device ,
The processing unit generates a void detection image indicating the detected void from the void candidate image based on the reference feature amount of the void,
When the filling rate of the void candidate image indicated by {(area of the void candidate image) / (area of the circumscribed rectangle of the void candidate image)} is higher than the reference filling rate as the reference feature amount, The image is determined as a void detection image, and when the filling rate of the void candidate image is equal to or less than the reference filling rate, the void candidate image is divided into void candidate element images, and the shape feature amount of the void candidate element image is calculated. And an inspection apparatus that determines the void candidate element image as a void detection image when the shape feature amount satisfies a void criterion . The inspection apparatus according to claim 6 , wherein
The processing unit is
An inspection apparatus that excludes from the void candidate images those existing in a predetermined region corresponding to some constituents of the subject. The inspection apparatus according to claim 7 ,
The processing unit is
An edge of the component is detected from an image obtained from the binarization processing of the X-ray image, a straight line region is set from the detected edge, and one of the void detection images determined based on the straight line region is set. Inspection device that masks parts. The inspection apparatus according to any one of claims 6 to 8 ,
The processing unit is
An inspection apparatus that controls the positions of the X-ray generator and the X-ray camera so that when a different part of the subject is imaged, the specific part of the subject becomes the center of the X-ray image.